Supabase CEO: databases under management jumped from 8 million to nearly 50 million in a year

Supabase's growth accelerates as agents, not just developers, adopt it

Paul Copplestone, co-founder and CEO of Supabase, told the hosts that signups grew from roughly 3 million a year ago to 13 million now, with revenue set to grow 4x this year. Databases under management have gone from about 8 million to nearly 50 million, with more than a million new databases created every week. He attributes the inflection to two forces: "people are building with abandon" among software engineers, and a wave of non-engineers now able to build on Supabase.

Copplestone said Supabase benefited from becoming a default database choice inside AI coding tools and from being open source, which he believes agents and enterprises both favor. "I think it was critical," he said of open source, noting Supabase also helps maintain the broader Postgres ecosystem rather than just its own tooling.

From dashboard to API to "code-first": three eras of the product

Copplestone described three phases: starting as a dashboard company built on Postgres (inspired partly by Airtable), shifting to an API-first company around 2021 as integrators needed to spin up databases programmatically, and now moving to a code-first architecture because, as he put it, "no one visits dashboards anymore." This year he issued an internal memo stating that everything in the dashboard must also work via CLI and infrastructure-as-code, so agents can operate Supabase directly.

"Builders" are displacing the assumption that you need a technical co-founder

Copplestone and the hosts discussed a shift from the 2022–2025 era, when AI's arrival made technical co-founders seem essential, to a newer view that domain experts without engineering backgrounds can now build sophisticated software themselves. He cited a taxi driver in Brazil who built a full taxi app on Supabase, and the hosts described a fintech founder building custom personal finance software with bank integrations from scratch.

"They've got all these ideas that they're churning out... I don't know where the ceiling is on this one." — Paul Copplestone

Personal databases as the future of individual data ownership

Copplestone discussed building his own personal software stack on Supabase—covering email triage, finances, health tracking, and file uploads—arguing that personal context stored in one database lets an agent correlate things like travel and weight fluctuations. He said he is now deliberately separating his "CRM" layer from any single agent so he can switch agents while keeping control of the underlying data.

"I certainly think that it'll be the case that everyone in the world with access to AI will become a builder of some sort." — Paul Copplestone

He framed Supabase's ambition in terms of scale: if platforms like ChatGPT have 1.2 billion weekly active users and each launched one database a month, that implies 1.2 billion databases monthly, a scale he said has never been required of database infrastructure before. To address this, Supabase acquired Turso, giving it a SQLite-compatible tier ("Supabase Spark") alongside a Postgres-scale product (Multigress) aimed at petabyte workloads.

Internal AI adoption and agent-oriented engineering

Copplestone said internal use of coding agents rose from 10% of engineers writing 80% of their code via agents to 60% within six months, driven by improving models and no formal internal AI policy. He said this has pushed heavy investment into testing and evals rather than just building, and that the company rewrote error messages specifically so agents can self-correct instead of abandoning the product when they get stuck.

Guest perspective: building a personal "brain" on Postgres

A guest described building a personal retrieval system ("G brain") on top of Supabase, using Postgres with PG vector for a graph-like structure over roughly 30–40 gigabytes of personal markdown data, arguing that large context windows alone aren't enough and that retrieval problems require structured databases. He described running continuous "captain" agents that auto-research ways to improve retrieval evals and ship improvements automatically, calling it "a very very primitive version of RSI for a particular use case."

Outlook: optimism tempered by open questions on team scaling and market shifts

Copplestone said he is "pretty pumped about 2027," citing rising entrepreneurship enabled by builders, though he conditioned this on avoiding "a few crazy events." The hosts and Copplestone noted vertical SaaS startups are struggling to raise Series A funding even at $5–10 million ARR, while small teams are reaching $10–50 million ARR with far fewer people and high margins—a shift Copplestone sees as largely positive for founders who no longer need to raise large rounds. He offered no specific numbers on how Supabase itself plans to restructure for future inflection points, saying only that dogfooding tools internally and maintaining spare team capacity for experimentation are central to staying adaptable.